Health: capacity and full range are over-estimated when a charge session is first seen mid-way #114

Problem

chargeSessions() builds each session from the first and last charging samples the app observed. session(start, last) (HealthSummary.kt#L164-L173)`` then mixes two kinds of value:

• startPercent = start.batteryLevel comes from the first observed sample;
• energyAddedKwh = last.chargeEnergyAdded and milesAddedRated = last.chargeMilesAddedRated are what the car reports cumulatively since the session began.

The first observed sample is often not the start of the session: the app was opened or came into range mid-charge, the link dropped and reconnected, or the phone was away during a scheduled overnight charge. In all of those cases, the energy and miles from the unobserved part get divided by only the observed SOC swing.

Example: the car is plugged in at 20 %. The first sample is at 50 % with charge_energy_added = 22.5, the last at 80 % with 45.0. Today's code computes 45 kWh / 30 % = 150 kWh. fullRangeMiles (cumulative miles / 30 %) is inflated by the same factor, and so is the "backbone" capacity kwhPerMile × fullRangeMiles. Using deltas gives (45 − 22.5) / 30 % = 75 kWh.

HealthSummaryTest always starts a session at energyAddedKwh = 0f, so this path has no test.

Two smaller problems in the same file:

• LEARNING_SESSIONS = 3 sits under the comment "The benchmark product needs about twenty sessions for a verdict; we do too." (L200-L201).`` The comment and the code disagree; one of them should change.
• Sessions are split purely on chargingState == "Charging". A brief Starting or Stopped state therefore turns one plug-in into several "sessions" and inflates the count used by the learning gate.

Proposal

• Use last − first deltas for energy and rated miles, or reject sessions whose first observed sample already has charge_energy_added > ε. Record the real session start when it can be observed.
• Add tests for: a session first seen mid-charge, a session interrupted by a disconnect, and Starting/Stopped blips.
• Move session segmentation into core/health next to the estimators; it is pure Kotlin.